1998
Zhu and Newell, Monthly Weather Review
Named atmospheric rivers and showed that a few narrow filaments carry most of the poleward water vapor transport in the midlatitudes.
Why this site exists
Demonstration case NP-01Integrated vapor transport · ribbons follow the flow
Forecast How a river grows
Form, move, intensify, land, fade
Five frames from the same run. The outline is the 250 kg m⁻¹ s⁻¹ contour, the line every detection method starts from. Watch the tail stay anchored near Hawaiʻi while the head runs northeast and the core fills in toward the coast.
Forecast Landfall
When and where it comes ashore
Each row is a point about 90 km off the coast. Read across for timing and up for place. The dark line marks 250 kg m⁻¹ s⁻¹, the threshold for AR conditions. The strip on the right ranks each latitude on the AR scale from its peak IVT and how long it stayed above 250.
By coastal point
The strongest flux comes ashore south of where the head first touches land. The head reaches Cape Mendocino first, but the core lands between Monterey Bay and Bodega Bay and sits there for two days as the river stalls.
Forecast Ensemble
How sure is the landfall?
Fifty-one members, each with its own timing, landfall latitude and strength. Pick a coastal point. The plume shows every member's IVT; the strip below gives the share of members above each AR threshold, six hours at a time.
Chance of AR conditions (IVT ≥ 250)
96%
Chance of AR3 or stronger
40%
Median onset · median peak
F060 · F072
Members by AR scale rank
Classify AR scale
Rank a river by strength and staying power
The AR scale of Ralph and others (2019) sets a preliminary rank from the peak IVT at a place, then moves it up one for 48 hours or more of AR conditions and down one for less than 24. Short weak events fall below the scale. Most AR1 and AR2 events refill reservoirs; AR4 and AR5 events bring floods.
AR3
StrongBalance of beneficial and hazardous
Primarily beneficialPrimarily hazardous
Load a coastal point from NP-01
Detect Tracking algorithms
Finding a river in a field of vapor
Every detector draws a contour around high transport, then tests the shape. Long and narrow passes; round blobs around cyclone centers and zonal bands of tropical moisture should not. Switch methods and move the setting to see which objects survive. Each method below runs live in your browser on this run.
How this method works
Diagnose Back-trajectories
Where the water comes from
Forty-five air parcels arrive off the North Coast at F072 at three levels. Traced back five days, they show the river's plumbing: low parcels load up over the warm subtropical ocean, while the 700 hPa parcels reach back into the tropics and rise along the warm conveyor belt in the final day and a half. Color is specific humidity along the way.
Pressure along the path
Where the arriving vapor was picked up
Uptake is booked when a parcel in the boundary layer gains moisture; rain-out along the way discounts earlier uptakes in proportion (after Sodemann and others, 2008).
Diagnose Intensification
Moisture or wind?
IVT is roughly the column's water vapor times the wind that carries it, so a river can strengthen by getting wetter or by getting faster. Following the core of NP-01 through the run, the curves of constant IVT show which one did the work.
AR core, every 6 hours
Change in peak IVT every 12 hours, split into moisture and wind
Methods Reading list
The literature behind each tool
2004
Ralph, Neiman and Wick, Monthly Weather Review
The first contour definition: water vapor of at least 2 cm in satellite imagery, longer than 2,000 km and narrower than 1,000 km.
Contour shape tests
2013
Wick and others, Weather and Forecasting
Automated the 2 cm contour into a detection tool and used it to score how well global models forecast landfalling rivers.
Forecast verification
2014
Rutz, Steenburgh and Ralph, Monthly Weather Review
Moved detection from water vapor to vapor transport: IVT of at least 250 kg m⁻¹ s⁻¹ along a feature at least 2,000 km long.
Fixed threshold tab
2015
Guan and Waliser, Journal of Geophysical Research
A global detector with a local 85th percentile threshold and tests for length, length-to-width, coherent direction, poleward transport and orientation.
Percentile tab
2018
Shields and others, Geoscientific Model Development
ARTMIP, the intercomparison project showing how much AR counts and climatologies depend on the choice of detector.
Why we show three methods
2019
Ralph and others, Bulletin of the AMS
The AR scale, AR1 to AR5, from peak IVT and the duration of AR conditions at a place.
AR scale and landfall ranks
2020
Xu, Ma, Chang and Wang, Geoscientific Model Development
IPART: image-processing detection by top-hat reconstruction of IVT, an axis found as the path of greatest flux through the object, and tracking by a modified Hausdorff distance.
Image-processing tab and axis
2008
Sodemann and others, Journal of Geophysical Research
Attributes moisture to its evaporation source along back-trajectories, discounting uptakes that later rain out.
Moisture sources
About this demonstration. Every map and chart on this page runs on NP-01, an analytic event built in the browser to behave like a landfalling North Pacific river: a subtropical tail, a lengthening axis, a strengthening low-level jet and a stall on the coast. The detectors, shape statistics, AR scale, ensemble probabilities, back-trajectories and moisture budget are computed live from those fields. The image-processing tab uses a plain top-hat (opening) on one time step with an anomaly floor of 50 kg m⁻¹ s⁻¹, a simplification of IPART's three-dimensional reconstruction. In production the same code reads IVT, water vapor and winds from the global models and ensembles on each cycle.